Latest 25 news items

Congrats to Harvey but we covered that already . AI News for 9/8/2026-9/9/2026.

Another swarm of AI agents in the wild, this time on a German-language forum, found by safety researchers looking for activity similar to the swarm that attacked Hugging Face. A couple of notes while reading the report at 1.

AI is rewriting the world and, at the same time, inventing a whole new language to describe how it’s doing it. Sit in on any product meeting, pitch, or panel these days, and you’ll hear people toss around LLMs, RAG, RLHF — and, as of last week, terms like “opaque recurrence,” the reasoning technique in OpenAI’s new Astra model that’s got AI safety…

In March last year, the Department of Cognitive Science at the University of California, San Diego, published the results of an experiment . It had asked almost 300 people to have a conversation with two mystery partners, one human and one AI, and try to determine which was which.

This guide is a fully public, inexpensive recipe for making a small model substantially better at structured-output compliance. We fine-tune LFM2.5-350M with Group Relative Policy Optimization (GRPO) using the TRL library and evaluate it on the IFStruct benchmark .

Launch season continues from yesterday , with Gemini 3.8 Flash as rumored today, but Muse Spark 1.3, promised in Zuck’s big comeback letter last month, definitely deserved the title story win today. Per AAII it is now the #3 model in the world (!?!) Just look at the confidence displayed finally putting up comparable numbers to the frontier models from…

For the entirety of the history of Generative Media , you basically had to design around the inconvenient fact that generating images and video takes time — even if you used consistency models to get a 30 second generation down to 1 second, you still only have a 1 FPS video at best… well below anything acceptable for consumer-grade human attention. Fal…

A late entrant in the news cycle of an eventful week: Following the closing of Cursor’s acquisition by SpaceX last week , it was time for OpenAI to do what Anthropic did to Windsurf when it was being considered for acquisition by OpenAI: We’re ending our partnership with Cursor following its acquisition by SpaceX. Under our proposal, Cursor’s direct access…

A technical walkthrough of how we built the Granite 4.2 reasoning model family. Authors: Granite Team, IBM TL;DR: Granite 4.2 is our first family of dense, decoder-only reasoning LLMs, released in three sizes: 3B, 8B, and 30B .

We’ve lost count of how many adoption milestones have been passed since the original Rise of the AI Engineer post, but surely Andrew Ng, cofounder of Google Brain and Coursera among many other things, relaunching DeepLearning.ai with a focus on AI Engineering is a big one : This was done via “ an analysis of over 10,000 job postings; carrying out dozens of…

By AI standards today is a pretty quiet Friday, so it’s time to take a step back and reflect on what is really going on. If you read our 2025 reading list , and followed our coverage of Z.ai GLM , understood the Poolside pivot , been following our AI for Science themes , and tuned in to today’s Simile pod , you not only are one of the biggest readers of…

Today, we release DSpark draft model checkpoints for three models from our LFM2.5 family: LFM2.5-1.2B-Instruct, LFM2.5-2.6B, and LFM2.5-8B-A1B. These add a speculative decoding path that trades a minimal memory increase for a large decoding speedup without changing output quality: Faster inference : up to 3.18 throughput improvement on a GPU and up to…

Last week was the 1 year anniversary of Zuck’s original Personal Superintelligence essay , and MSL seems to be feeling a second wind this year, as they slowly ramped up with the Dreamer acquisition and then Muse Spark and recently Muse Code . For a while it seemed like MSL was being rather timid with the launches… but today that all changed.

Meta is back with Muse Glimmer: local, agentic, multimodal, and open source! Great news from the OGs of open source LLMs!

Making Knowledge Distillation Cheap Enough to Run at Scale Knowledge distillation , training a smaller student model to match the performance of a larger teacher, is a well-known technique in Machine Learning. With the recent wave of open-source Large Language Models, such as gpt-oss , Qwen , GLM , or Kimi , it has become a mainstream research topic again.

[Submitted on 5 Aug 2026 ( v1 ), last revised 10 Aug 2026 (this version, v2)] Title: Latent Fact-Checking: Detecting Misinformation through Activation Engineering View a PDF of the paper titled Latent Fact-Checking: Detecting Misinformation through Activation Engineering, by Pedro Barcelos and 5 other authors View PDF HTML (experimental) Abstract: The…

We’ve discussed the HuggingFace-OpenAI security incident before, but OpenAI’s side of the story was the talk of the town at Black Hat (summaries from former guests Elie and Simon are worthwhile): At the core of OpenAI’s disclosures was how their models figured out how to use OpenAI’s internal Artifactory as a messageboard to orchestrate themselves:…

In The Custom ASIC Thesis we said Taalas was worth paying attention to, and in the Inference Inflection we said everything would go vertical. Our Baseten episode had some skeptical counterpoints against etched LLMs, not just custom ASICs , but clearly Lisa Su disagrees for now.

[Submitted on 20 May 2026] Title: Simulator-Grounded Large Language Models for Industrial Causal Reasoning: Tool-Use, Structured Injection, and Plant-Portable Retrieval for Wastewater Treatment Decision Support View a PDF of the paper titled Simulator-Grounded Large Language Models for Industrial Causal Reasoning: Tool-Use, Structured Injection, and…

Part of our Inference Engineering Masterclass pod yesterday involved a spicy discussion about Megakernels: megakernels are dead why are megakernels useful? you spend two months writing a kernel to save time on launch overhead and poor inter-kernel overlap.

Deploy local agents everywhere with LFM2.5-2.6B LFM2.5-2.6B is built to power capable agents entirely on-device. It supports tool calling and multi-step workflows while staying small and fast enough for everyday hardware, from laptops to phones.

After the Qwen Exodus last year and new management took over launching more closed model APIs, there was some real doubt as to whether or not this leading open models lab would continue to release relevant models. That doubt is now gone.

We first covered Baseten last year when DeepSeek mania was at peak hype. Now they have raised a monster $13B round and become one of the new cohort of AI Infra decacorns that are (with Nvidia, Intel, and the semis complex) chief beneficiaries of the Inference Inflection .

When The Atlantic published a searchable dataset of works used to train AI, Kirk Wallace Johnson, like a lot of artists, looked for his name out of curiosity. And, like a lot of artists, he found it.

[Submitted on 21 Jul 2026 ( v1 ), last revised 28 Jul 2026 (this version, v2)] Title: CausalGate: Causal Importance Distillation for Transformer Module Pruning View a PDF of the paper titled CausalGate: Causal Importance Distillation for Transformer Module Pruning, by Kiran Nair and 2 other authors View PDF HTML (experimental) Abstract: Existing adaptive…